Abstract

This dissertation investigates methods for the automated design and optimization of laminated composite structures. Optimal design of laminated composite structures is challenging due to the possibility to locally adapt the material system to the mechanical situation. Automated structural design on a computer is enabled by a combination of numerical simulation and optimization algorithms. The finite element method provides the possibility to predict mechanical properties of virtual candidate solutions. Numerical optimization algorithms then adapt the structure’s attributes in order to meet specific demands formulated on the aforementioned simulated properties. Evolutionary algorithms are a group of biologically inspired optimization algorithms which have repeatedly and successfully been applied to optimal design problems with laminated composites. This thesis focuses on methods to compose evolutionary algorithms for the specific traits of laminate optimization problems. A special focus is set on the variation state of a canonical evolutionary algorithm. This state is particularly influenced by the genetic representation of a candidate solution, i.e. the way the adjustable attributes are translated to machine readable entities. The aim of the thesis is to develop and examine genetic representation schemes to concurrently evolve a structure’s topology, shape, and laminate properties. An overview of structural optimization and evolutionary computation illustrates the state-of-the-art. In variable-length representations, the dimensionality of the search space is a variable to optimize. The importance of variable-length representations in evolutionary topology and laminate optimization is exemplified. A weakness of established variable-length crossover operators is the treatment of length constraints. Based on existing concepts a split-and-splice variable-length crossover operator respecting length constraints is introduced. In order to improve the solution quality of a real-encoded evolutionary algorithm, a gradient-based local search is embedded in the variation state of the algorithm. The algorithm intrinsic parallelization is extended to the variation state in order to cope with different runtimes of deterministic and stochastic operators. A parallelization of the variation state requires abandoning of synchronization points. Hence, the population is replaced by a pool of individuals where distributed breeder processes continuously draw samples for mating and insert offspring to replace parents. A lifetime concept is developed to keep the pool size approximately constant. A niching strategy focuses the stochastic component of the algorithm to unexplored regions

Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call

Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.